Update Dockerfile

Dockerfile

@@ -1,39 +1,43 @@
-FROM ghcr.io/ggml-org/llama.cpp:full
-
-WORKDIR /app
-
-RUN apt update && apt install -y python3 python3-pip python3-venv wget && rm -rf /var/lib/apt/lists/*
-
-RUN python3 -m venv /opt/venv
-ENV PATH="/opt/venv/bin:$PATH"
-RUN pip install -U pip huggingface_hub
-
-# Define the build argument for the token
-ARG HF_TOKEN
-
-# Download Gemma 4 26B using the token to prevent the warning and rate-limiting
-RUN python3 -c 'from huggingface_hub import hf_hub_download; \
-    import os; \
-    token = os.getenv("HF_TOKEN") or None; \
-    repo="unsloth/gemma-4-26B-A4B-it-GGUF"; \
-    hf_hub_download(repo_id=repo, filename="gemma-4-26B-A4B-it-UD-Q4_K_XL.gguf", local_dir="/app", token=token); \
-    hf_hub_download(repo_id=repo, filename="mmproj-BF16.gguf", local_dir="/app", token=token)'
-
-# Download Gemma 4 E2B + Speculative Setup
-RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf" -O /app/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf && \
-    wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/mtp-gemma-4-E2B-it.gguf" -O /app/mtp-gemma-4-E2B-it.gguf && \
-    wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/mmproj-F16.gguf" -O /app/gemma-4-E2B-it-mmproj.gguf
-
-# Download Gemma 3 1B Model
-RUN wget "https://huggingface.co/MaziyarPanahi/gemma-3-1b-it-GGUF/resolve/main/gemma-3-1b-it.Q3_K_M.gguf" -O /app/gemma-3-1b-it.Q3_K_M.gguf
-
-CMD ["--server", \
-     "-m", "/app/gemma-4-26B-A4B-it-UD-Q4_K_XL.gguf", \
-     "--mmproj", "/app/mmproj-BF16.gguf", \
-     "--host", "0.0.0.0", \
-     "--port", "7860", \
-     "-t", "2", \
-     "--cache-type-k", "q8_0", \
-     "--cache-type-v", "iq4_nl", \
-     "-c", "64000", \
-     "-n", "38912"]
+FROM ghcr.io/ggml-org/llama.cpp:server
+
+RUN apt update && apt install wget -y && rm -rf /var/lib/apt/lists/*
+
+# Create a dedicated directory to organize the models
+RUN mkdir -p /models
+
+# --- MODEL 1: Gemma 4 Multimodal + Speculative Setup ---
+RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf" -O /models/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf
+RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/mtp-gemma-4-E2B-it.gguf" -O /models/mtp-gemma-4-E2B-it.gguf
+RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/mmproj-F16.gguf" -O /models/gemma-4-E2B-it-mmproj.gguf
+
+# --- MODEL 2: Gemma 3 1B Model (Corrected Direct Download Link) ---
+RUN wget "https://huggingface.co/MaziyarPanahi/gemma-3-1b-it-GGUF/resolve/main/gemma-3-1b-it.Q3_K_M.gguf" -O /models/gemma-3-1b-it.Q3_K_M.gguf
+
+
+# --- CREATE ROUTER CONFIGURATION ---
+# Appending configuration step-by-step to avoid any multiline shell escaping issues
+RUN echo "[gemma-4-vision]" > /models.ini && \
+    echo "model = /models/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf" >> /models.ini && \
+    echo "spec-draft-model = /models/mtp-gemma-4-E2B-it.gguf" >> /models.ini && \
+    echo "mmproj = /models/gemma-4-E2B-it-mmproj.gguf" >> /models.ini && \
+    echo "ctx-size = 4096" >> /models.ini && \
+    echo "flash-attn = on" >> /models.ini && \
+    echo "ubatch-size = 128" >> /models.ini && \
+    echo "batch-size = 512" >> /models.ini && \
+    echo "spec-type = draft-mtp" >> /models.ini && \
+    echo "spec-draft-n-max = 3" >> /models.ini && \
+    echo "" >> /models.ini && \
+    echo "[gemma-3-1b]" >> /models.ini && \
+    echo "model = /models/gemma-3-1b-it.Q3_K_M.gguf" >> /models.ini && \
+    echo "ctx-size = 4096" >> /models.ini && \
+    echo "flash-attn = on" >> /models.ini
+
+
+# --- START SERVER IN ROUTER MODE ---
+CMD [ \
+  "--models-preset", "/models.ini", \
+  "--port", "7860", \
+  "--host", "0.0.0.0", \
+  "-t", "2", \
+  "-tb", "2" \
+]